bytedance/deer-flow

An open-source long-horizon SuperAgent harness that researches, codes, and creates. With the help of sandboxes, memories, tools, skill, subagents and message gateway, it handles different levels of tasks that could take minutes to hours.

What it solves

DeerFlow is a "super agent harness" designed to orchestrate complex research and task execution. It solves the problem of managing multiple sub-agents, long-term memory, and secure code execution environments (sandboxes) within a single framework, allowing it to perform a wide variety of autonomous tasks.

How it works

DeerFlow acts as a central orchestrator that manages the following components:

  • Sub-Agents: It can spawn and coordinate specialized agents to handle different parts of a task.
  • Skills & Tools: It uses extensible skills and integrates with MCP (Model Context Protocol) servers to expand its functional capabilities.
  • Sandboxes: It provides isolated environments (Local, Docker, or Kubernetes) where the agent can safely execute code and manage a file system.
  • Memory: It implements long-term memory and context engineering (including manual context compaction) to maintain coherence over long sessions.
  • Session Goals: It uses goal-oriented tracking to ensure the agent stays on task.

Who it’s for

It is intended for developers and researchers who want to build and deploy sophisticated AI agents capable of deep research, autonomous coding, and complex workflow automation.

Highlights

  • Flexible Deployment: Supports local development, Docker, and Kubernetes production environments.
  • Extensible Capabilities: Integrates with Claude Code, MCP servers, and a specialized intelligent search toolset called InfoQuest.
  • Comprehensive Tooling: Includes a Terminal Workbench (TUI) and a sister project, LLM Space, for prototyping and debugging agent steps.
  • Model Agnostic: Compatible with various LLM providers including OpenAI, OpenRouter, vLLM, and specific ByteDance models like Doubao-Seed-2.0-Code.